MERGER LAB

An M&A accretion/dilution engine for Indian public markets — build the model, ship the IC memo.

Live deal room: doginfantry-merger-lab.hf.space · Hugging Face Space

Why now

RBI's acquisition-finance liberalization (amended 13 Feb 2026, effective 1 Apr 2026) allows Indian banks to finance corporate acquisitions for the first time: up to 75% of acquisition value in bank debt, minimum 25% acquirer equity, maximum 3:1 consolidated debt-to-equity, acquirer net worth ≥ ₹500 crore with a 3-year profitability track record. That opens a structural M&A/consolidation wave in India — and every deal in that wave needs exactly the analysis this engine produces.

What it produces

1. An IC memo PDF — pyramid-principle, recommendation first, with a regulatory-stack section (RBI financing compliance, SEBI SAST open-offer scenarios, MPS/CCI flags) no generic template has.
2. A linked Excel model — every tab downstream of Assumptions is live cross-referenced formulas; a Δ-vs-engine column proves the workbook ties to the Python engine to the last decimal.
3. A Streamlit deal room (secondary — the dashboard is a wrapper, the memo and Excel are the product).

Three pre-run sample deal rooms live in samples/ — including an honest DECLINE: when a 24x-earnings acquirer buys a 60x-earnings target, the engine says so.

Sample	Structure	Verdict
Project Horizon (IT)	100% acquisition, 60% stock, collar priced	DECLINE — mechanical dilution
Project Bastion (Cement)	64% promoter block + 26% open offer	DECLINE — premium unsupported
Project Meridian (Metals)	51% + open offer, 100% cash in RBI guardrails	PROCEED WITH CONDITIONS

Module map

Module	What it does
src/data_layer.py	yfinance + Screener.in CSV fallback, INR-crore normalization, 24h cache, USD-filer FX inference
src/precedent_db.py	SQLite precedent DB, 37 India deals 2019–2025; window-function premium percentiles, CTE comps — raw SQL on purpose
src/sources_uses.py · deal.py · ppa.py	S&U balanced to the rupee (asserted), PPA with DTL and goodwill
src/rbi_compliance.py	The five RBI 2026 guardrail checks, PASS/FAIL with explanations
src/sebi_sast.py	25% trigger → 26% open offer, acceptance scenarios, MPS breach, creeping acquisition, CCI threshold
src/accretion_dilution.py	Y1–3 pro-forma EPS engine, analytic break-even synergies, earnings-yield heuristic cross-check
src/contribution.py · sensitivity.py	Contribution vs ownership exhibit; two-way grids that re-run the full engine per cell
src/optimizer.py	SLSQP financing-mix optimizer against the real engine, binding RBI constraint named in the memo
src/monte_carlo.py	10,000 seeded iterations; P(accretive by Y2), P5/P50/P95
src/value_bridge.py	Synergy PV vs control premium, ROIC vs WACC, "accretive but value-destructive" warning
src/collar.py · merger_arb.py	Black-Scholes exchange-ratio collar (long put / short call), market-implied close probability
src/memo_generator.py · excel_generator.py	The deliverables: Jinja2→PDF memo with inline-SVG charts, 10-tab formula-linked workbook

Data sources — zero paid APIs

yfinance (prices, financials, shares), manual Screener.in CSV exports (fundamentals fallback), curated precedent transactions from SEBI SAST letters of offer and NSE/BSE filings. Debt cost anchored to RBI DBIE lending-rate publications; risk-free = 10Y G-Sec (CCIL/FIMMDA). No paid APIs, no scraping, no LLM calls anywhere in the pipeline.

Methodology notes

• Combined NI = acquirer NI + owned% × target NI + after-tax phased synergies − after-tax new interest − after-tax foregone cash yield − after-tax incremental D&A − after-tax Y1 integration costs.
• Break-even synergies solved analytically (the equation is linear).
• Every module carries a methodology docstring; every simplification is documented where it lives.
• tests/test_known_deal.py is the credibility anchor: a toy deal with every expected value derived by hand in the comments, asserted to 0.1%. 28 tests total.

How to run

pip install -r requirements.txt
python tests/test_known_deal.py        # the sacred hand-checked suite
python generate_samples.py             # rebuild the sample deal rooms (live yfinance data)
streamlit run app/streamlit_app.py     # the deal room

The landing page in site/ is a single static file — deploy with vercel site/ (set the Streamlit URL in index.html first).

Limitations & disclaimers

Illustrative analysis on public data for portfolio demonstration — not investment advice. Sample transactions are hypothetical. Precedent-DB rows marked ILLUSTRATIVE — verify carry unverified numbers pending checks against the cited filings. Standalone EPS held flat Y1–3; partial stakes consolidated economically (no minority-interest line); SAST floor price modeled at the deal offer price (real floor is the highest-of tests incl. 60-day VWAP). Verify RBI/SEBI parameters against the master directions before relying on any output.

Roadmap

• Verify precedent seed rows against letters of offer (replace ILLUSTRATIVE tags)
• Regulatory WACC helper (CAPM with India ERP input) and target-price DCF cross-check
• Deal-financing term sheet exhibit (tenor/amortization schedule vs RBI ongoing D/E test)
• Hindi-language memo variant

---

MERGER LAB · DogInfantry · repo: M&A accretion/dilution engine for India — IC memo + Excel model outputs. RBI 2026 acquisition-finance framework built in.